Suppose, for example, that a bar of steel is being investigated; it can be put into a machine and pulled until it breaks in two. The machine registers the amount of the pull which was applied. Or a small piece can be put under a press and compressed to any desired degree. It can also be tested by impact or even pulled apart by a sudden blow, as described in _Mechanical Inventions of To-day_. The bar can be supported by its ends and loaded or pulled down in the centre, so that its power of resisting bending can be determined. It can be judged, too, from its chemical composition. Steel, in particular, depends for its properties very largely upon its chemical composition. The difference between cast-iron, wrought-iron and steel, also the differences between the innumerable varieties of steel, are due almost entirely to the admixture of a certain percentage of carbon with the metal. This can be ascertained by chemical analysis. This form of inquiry has the advantage over the more purely mechanical methods in that the latter, for the most part, have to be applied to the bar as a whole, whereas the quality may vary in different parts, the surface in particular being liable to differ from the interior. In such cases, one analysis can be made of a piece cut from the surface and another of a piece from the centre. And it is here, too, that microscopical analysis comes in. For this purpose a piece is sawn off the bar, and the end ground perfectly smooth. This is then washed in a suitable chemical, such as a mild acid, which acts differently upon the different materials of which the "metal" is built up, thereby rendering them visible one from another. A photograph taken through a microscope then shows the structure of the metal; how the different constituents are built together. This is known as metallographic testing, and its advantage as compared with chemical analysis is that the latter shows, as we might say, what are the bricks of which the thing is built, while the former shows how the bricks are arranged. Indeed it is hardly correct to speak of the advantage or superiority of one over the other, since each is the complement of the other, supplying the information which the other fails to give. And there are other mechanical tests which have not yet been mentioned. There are machines which twist a bar so as to discover its power to resist torsion, there are others which apply a downward pressure on one part of the bar and an upward one on an adjacent part, so as to show its capabilities in withstanding shearing strain.